Challenging Glass 5 – Conference on Architectural and Structural Applications of Glass

Belis, Bos & Louter (Eds.), Ghent University, June 2016.

Copyright © with the authors. All rights reserved.
ISBN 978-90-825-2680-6

Manchester Town Hall, a Case Study in Structural Glass

Reliability and Robustness
B. Eckersley a, G. Coult a & P. Lenk b
a
Eckersley O’Callaghan, UK, graham@eocengineers.com
b
Formerly of Eckersley O’Callaghan, UK

The design of the structural glass for the Manchester Town Hall Link (completed 2015) was carried out by engineering
consultancy Eckersley O’Callaghan (EOC). The glass facade forms an enclosed shell which supports a steel roof and
acts monolithically to resist lateral loads. The project collates the latest advances in glass technology combined with
innovative design methods. In the absence of explicit codes of practice for structural glass, EOC performed a first
principles approach in using empirical data, acquired through previous projects, and analytical methods, including
FEM and parametric modelling, to justify an elegant and efficient structure which was approved by building control
authorities. The result has reduced glass joints and less visible metalwork to meet the architects’ aspirations. This
paper describes the approaches and innovations in designing the structural glass for this project.

Keywords: Glass, Robustness, Buckling, Material Properties

1. Introduction
The Library Walk pavilion is a link between the remodelled Grade II-listed extension building of the Manchester
Town Hall and the adjacent Central Library, designed by Simpson Haugh and Partners.

The 175m2 pavilion uses frameless structural glass panels to support a 30-tonne, stainless steel roof structure. The
distinctive shape of the roof was form-found using mathematical algorithms designed to create a smooth, organic
but ‘rational’ undulation in the soffit, based on spherical distortions of a flat surface. This undulation responds to the
arched windows of both extensions.

Fig. 1 Manchester Town Hall Link Building

The roof consists of polished stainless steel monocoque construction consisting of top and bottom doubly-curved
structural surfaces contributing to the stiffness and strength of the roof, allowing it to span 15m across a column-free
space. These external exposed surfaces are welded to an internal armature of stiffeners, creating a rigid structure.
This monocoque construction was engineered and fabricated by specialist manufacturers, CIG, using automated
bending techniques combined with a manual crafted polishing of the surface. Large segments were prefabricated in
the factory and bolted, welded and polished on site at the seams.
 Challenging Glass 5
The structural glass façade collates the very latest in glass technology coupled with cutting-edge analysis and design
to refine the visual simplicity and transparency. As there are no vertical fins, full-height glass panels are utilised; this
innovative design has fewer glass joints and less visible metalwork to exceed the architect’s aspirations. The façade
includes both straight and curved panels. The 7.2m tall, base supported panels support the roof and provide lateral
stability to the whole pavilion.

The simplicity and purity of this building is achieved by the simple combination of the two structural elements of
the roof and the vertical glass façade, which are rigid in virtue of their form. In other words, the structure is the form,
the enclosure as well as the supports.

Library Walk was an opportunity to experiment with new design tools, technologies and fabrication methods to
allow the pure expression of structure in sculptural forms. In both the roof monocoque, delivered using innovative
method of fabrication borrowed from the shipbuilding industry and refined to new levels of precision, and in the
structural glass façade, taking the latest in reductive engineering to its zenith; form, enclosure and structure have
been merged into a single integrated system.

As well as being a playful, sophisticated and sensitive sculptural piece, the project opens new opportunities within a
more integrated approach to the design and fabrication of structural and cladding systems.

This paper will discuss the design approach adopted to ensure robustness and reliability absence of explicit codes of
practice.

1.1. Company Introduction

Eckersley O’Callaghan (EOC) is recognized as one of the leading authorities in the structural design of glass
through their notably high profile work with Apple Inc. amongst others. These structural glass projects are
frequently referenced as the cutting edge in both glass design and ambition. Further, and perhaps most relevantly,
these structures have been designed and justified in many different countries through many different building
control authorities. EOC’s expertise is therefore more than just the successful design of these glass structures, it is
the successful track record they have had in obtaining permissions for over 40 major glass structures in over 10
different countries (including the UK) and cultures in the last 10 years. Specifically, in the UK EOC has designed
and had approval for glass structures in London, Bristol, Exeter, Milton Keynes, Glasgow, Belfast, Liverpool and
Manchester.

1.2. Eckersley O’Callaghan’s Role

Eckersley O’Callaghan were the structural engineers for the detailed design of the façade. Both the roof and
substructure were designed by others. The glass panels act as the interface between the two structures; careful co-
ordination of both deflections and load transfer paths were required throughout the design process to ensure a
cohesive structural system.

2. Structural System

2.1. Gravity System

The roof structure is supported on 20 laminated glass wall panels. The panels are constructed from 3 fully-tempered
12 mm glass lites, laminated with 1.52 mm SGP interlayers. The geometry of the building was optimized from the
design intent resulting in a simpler form that is divided into regions of flat glass panels and curved glass panels of 2
different radii with negligible geometric difference. Flat glass panels are stiffened by perpendicular glass panels,
which serve to connect the new glass building to the two existing buildings. There are 6 sliding glass doors within
the façade. Portal steel frames around the doors are design to carry the weight glass panels above the door and
provide lateral restraint to the neighbouring glass panels. To minimize the door frame steel size, the roof does not
transfer load to the glass panels over the door frames, instead the roof spans over the openings and loads the glass
panels as shown in figure 2b.
 Manchester Town Hall, A case study in structural glass reliability and robustness

a) b) Global Analysis Model

Fig. 2 Gravity Load Resisting System a), Panelization b) Global Analysis Model

Each glass panel is base supported within a semi-rigid steel shoe, and restrained laterally at roof level. Gravity loads
are transferred to the steel shoe by a central bearing block, as shown in figure 3. The steel shoe is supported on a
reinforced concrete ring beam at the panel corners. Roof loads are transferred to the glass panel via a central bearing
bloc only.

b)

a)
Fig. 3a) Panel Loading and Support Points, and b) Semi-rigid shoe

2.2. Lateral System

The global stability is primarily provided via shear transfer at vertical silicone joints between glass panels (Figure
4a). The roof acts as a stiff diaphragm connecting all glass panels together; the bonded glass walls then form a stiff
loop which resists overturning moment by the second moment of inertia of the entire building. This significantly
reduces internal glass stresses from stability loads.

A secondary stability system is also provided by silicone in the base shoe connection detail (Figure 4b). This system
resists panel overturning by a push-pull action at panel corners. This system acts as a secondary, redundant system
to the primary vertical silicone joint system, and is utilised in the event case of failure a glass panel.
 Challenging Glass 5

b)

a)
Fig. 4a) Primary Stability System, and b) Secondary Stability System.

2.3. Design Approach

Our design approach is based on a combination of published codes, in house testing data, published papers and
journals, research and development, prior project experience and test results, published guidelines where they
remain current and the design experience developed from over 20 years of active involvement in the field of
structural glass. Due to the lack of a valid European code for glass strength and its limitations in the application to
structural glass an approach based on the American ASTM E1300 (2012) was adopted for the design of glass
elements. Where relevant BS EN codes exist, (including the draft standard prEN 13474-3 (2008) and now published
IStructE guidance (2015)) their guidance was followed, and figures were justified with in-house testing when
necessary. In order to ensure that the Town Hall Link Building met the Eurocode-defined minimum level of
structural reliability, a first order reliability analysis was carried out. In order to ensure that the material data
assumed within the design process was matched by the finished construction a stringent set of quality control tests
was defined, and carried out by a third party.

Structural analysis was carried out using both standard analytical methods and numerical methods and checked
using simplified approaches. Panels were subdivided into 4 design groups: 2 flat and 2 curved. For each group the
critical panels were identified using the global analysis model.

3. Design Requirements

3.1. Collaboration at Interfaces

The glass panels span between a reinforced concrete slab at base level and the steel roof. Both the roof and the floor
were designed by others, and have very different stiffness properties. Calibration was required at both levels in order
to accurately predict the load distribution from the glass roof to the glass panels. An initial assumption of even
distribution of roof weight to each panel was taken, which provided an initial estimation for the spring stiffnesses at
the supports. The spring stiffness at the base was refined from deflection plots provided by URS. This provided a
more accurate distribution of roof loads within the glass panels.

Additionally, thermal differential movements between the roof and the RC slab needed to be accommodated within
the glass walls. Lateral expansion of the roof was accommodated by allowing the panels to tilt, rotating around the
minor axis. Internal temperature changes cause hogging and sagging in the steel roof, which in turn causes the
distribution of roof loads in the panels to vary.

3.2. Robustness and Redundancy

Redundancy has been provided for both the gravity and stability force resisting systems. In the event of accidental
damage, three different redundant design conditions were identified:

a) Failure of a single ply

b) Failure of critical glass panel
c) Failure of 2 critical glass panels

If a complete panel is broken, the roof is designed to span over the soft spot and redistribute forces to the remaining
glass panels. Additionally, the roof acts to tie the remaining glass panels together. The distribution of loads to the
remaining intact panels was assessed, as shown in figure 5 for redundancy conditions b and c. Reduced partial safety
factors were utilised for the accidental design.
 Manchester Town Hall, A case study in structural glass reliability and robustness

a) b)
Fig. 5a) Failure of critical panel within a design group, and b) Failure of two critical panels within each design group.

Global stability after accidental breakage was ensured by the secondary shear connection at the panel base, as shown
in figure 3.

3.3. Design Reliability

Eurocodes 1991 - 99 ensure that all Eurocode compliant structures have a minimum level of structural reliability.
Annex B and C of BS EN 1990 (2002) specify reliability in the form of a reliability index (β value). The defined β
values vary with building consequence class.

Partial factors for material safety and applied loads are defined within individual Eurocodes in order to achieve a
minimum reliability of β=3.8, at ultimate limit state for a 50-year life span. This is the β value associated with
medium consequence class buildings. In the absence of a valid European Code for structural glass, a global factor of
safety is required in order to meet the minimum structural reliability defined in BS EN 1990.

A global factor of safety for buckling has been calculated according to the rules established by the Joint Committee
on Structural Safety (JCSS). For a specific design condition comprising dead and snow load, a reliability assessment
was carried out. For this, a first order reliability method (FORM) was adopted through the use of calculation
software CodeCal developed by Faber et al. (2003). A Weibull distribution was taken for the glass material strength,
with a coefficient of variation of 15% for fully tempered glass that was to be corroborated by testing during
fabrication quality control. For the required load combination, a global factor of safety on buckling was calculated
as 1.8 at ULS, for the 50-year design life.

4. Analysis Approach
Three levels of numerical modelling were used to analyse the structural system. At the global scale fittings are
modelled as simple beam elements, façade panels as plate elements and fins as beam elements. These models use
appropriate techniques to reflect the relative stiffness of the structural system such that load paths are representative.
From this model the critical areas and the load path can be identified.

Identified critical panels, restraints and connections are subsequently modelled in greater detail by 2-dimensional
analysis models. After validation of these models by simple hand checks, detailed volumetric models of key fitting
connections or areas of interest are created.

For the two dimensional models, the complex behaviour of laminated glass is approximated with an effective
thickness approach. For detailed stresses, for example around fittings, full 3D volumetric models are used with the
bulk interlayer modelled.

4.1. Effective Thickness Modelling

Both analytical and numerical modelling techniques were used to establish the effective thickness of the three ply
build-up.

The propped cantilever arrangement was modelled as a 2D-plane strain model with the laminate build up as shown
in figure 5. The deflection under out of plane loading was calculated, from which the effective thickness of an
equivalent glass monolith was determined.
 Challenging Glass 5

a) b)
Fig. 6a) Failure of critical panel within a design group, and b) Failure of two critical panels within each design group.

Additionally, the effective thickness has been calculated using the approach presented in prEN13474-3 (2008) for
comparison. The lesser of the two effective thicknesses was used for design.

4.2. Buckling Analysis of Glass Panels

Non-linear finite element buckling analysis of both curved and flat glass panels was carried out using Strand 7.

Different combinations of initial fabrication imperfections, out of plane wind loading, and lateral roof movement
were investigated for each panel type. For a combination of initial bow imperfections of L/1000, wind induced
deflection of L/1000 and a horizontal roof movement of 3.6 mm (assessed based on the thermal expansion of the
roof), the buckling response is shown in figure 7.

a) b)
Fig. 7) Non-linear buckling analysis of flat glass panel, a) Glass stresses under vertical load for defined initial imperfections, b) Force
displacement response for critical node

In order to validate the non-linear analysis a series of increasing complexity models were produced. The theoretical
buckling capacity was calculated analytically assuming each panel to act as a propped cantilever with axial point
loads at the tip and mid-height.

A linear buckling analysis was then carried out in Strand 7 for both flat and curved panels. A variety of base support
conditions and initial imperfections were considered in order to calibrate the numerical model. The numerical model
was found to converge well to the theoretical solution. Figure 8 shows the analysis of a single flat panel. The applied
loads in figure 8b are extracted from the global analysis model shown in figure 2b. In figure 8a the buckling
response under different boundary conditions is shown alongside the theoretical buckling response.
 Manchester Town Hall, A case study in structural glass reliability and robustness

b)
a)
Fig. 8a) and b) Linear Buckling Analysis of flat glass panel

4.3. Detailed Sub-Modelling of Connections

Detailed modelling of connection elements was carried out in Strand 7. Volumetric models were constructed of key
elements including:

 Vertical silicone joints

 Base silicone connection
 Base shoe
 Polymeric bearing blocks at floor and roof level

The first two models were used to confirm the silicone stresses under lateral loading. The base shoe model was used
to determine the support conditions for the glass panels, and lastly the bearing block models were used to confirm
contact stresses within the glass panel.

A variety of joint shapes were experimented with for the design of the bearing blocks in order to optimise the
bearing stresses in the glass panel. A solution that utilized a tapered bearing block, as shown in figure 9a was found
to produce an even stress distribution within the glass panel, without large peak stresses.

a) b)
Fig. 9 Bearing block local model a) Tapered bearing block b) Glass stress distribution under gravity loading
 Challenging Glass 5
5. Material Properties
For all analysis, numerical or analytical, accurate solution depends on accurate material data. This requires material
knowledge of the glass, interlayer and silicone joints.

5.1. Glass Material Strength

In the absence of Eurocode defined material strength of glass, design data has been adopted from the American
standard ASTM E1300. Three different load durations have been specified and the corresponding material strengths
used in design.

5.2. Interlayer Material Properties

SGP is viscoelastic, a different shear modulus is required for different durations of loading. Design data is available
from Kuraray (then DuPont) (2014) for a variety of different temperatures and load durations. SGP bulk material
properties were acquired by Dynamic Mechanical Analysis (DMA) carried out for a previous project by an
independent body. The results were comparable to those published by Kuraray but enabled us to extrapolate beyond
the table of published values (see Figure 10).

c) d)
Fig. 10 SG5000 material properties a) DMA data b) Published Kuraray Data (2014)

5.3. Structural Silicone Material Properties

A Mooney – Rivlin material model was determined by dynamic mechanical analysis carried out for a previous
project. Material constants were calculated directly from test data within Strand 7 (Figure 11, Table 1)

Fig. 11 Curve Fitting to establish Mooney-Rivlin Material Properties

Table 1: Structural Silicone Material Properties.

3
Material Density [kN/m ] Young’s Poisson Ratio Tensile Strength Design Strength Thermal
Modulus [MPa] [MPa] [MPa] Expansion
[µm/m/K]
Silicone 0.98 1.4 0.499 1.2 0.138 9.2
 Manchester Town Hall, A case study in structural glass reliability and robustness

6. Validation and Quality Assurance

Design assumptions were validated through a combination of material testing and numerical modelling. Additionally,
a stringent set of quality control tests were specified and carried out to verify the data taken for analysis.

6.1. Validation of stability system.

In design it was assumed that the vertical silicone joints provided the primary stability system, with silicone in the
base shoe providing a secondary system in the event of panel failure. This assumption was tested by comparing the
relative stiffness of the two systems. Global analysis models of the structure were defined with 1) vertical silicone
joints only, 2) base silicone joints only, and 3) both sets of joints. The vertical system was found to be 30% more
stiff than the base joints (Figure 12). Both systems were found to be capable of resisting lateral movements
independently.

Fig. 12 Global lateral deflection of independent stability systems

6.2. Validation of silicone design assumptions

Non-linear shear properties of structural silicone are not yet well understood. In order to confirm the linear
behaviour of the structural silicone in shear, a combination of material test data, bulk finite element models and
published journal articles were used. Dias et al. (2012) and Bondi et al. (2009) have both shown that silicone non-
linear behaviour only occurs beyond strains of approximately 60%; volumetric finite element models of the
structural silicone joints, based on material properties established by DMA showed that for the geometry and
loading conditions considered here, strains of 13% were observed. The silicone is well within the linear range of
behaviour.

Additionally, project specific mock-ups of the base shoe were tested in tension as part of the quality control
procedures. Large scale pull-out tests, as shown in figure 13, were carried out. A tensile force was applied to the
glass-to-shoe connection, and the deformation and failure capacity recorded. The results were used to validate the
detailed volumetric finite element models of the base connection shown in figures 13e) and f) an also the curing of
the deep joint. Additionally, the finite element modelling assumed that silicone deformation and failure occurred
within the bulk silicone material and not the silicone-glass or silicone-steel interface. Pull out tests were performed
to failure, and the failure mechanism recorded. When adhesive failure occurred the joint preparation specification
was adjusted to ensure cohesive behaviour.
 Challenging Glass 5

a) b)

c) d)

e) f)
Fig. 13a) Schematic of base shoe pull-out test procedure, b) Test performed by independent testing authority, c) Cohesive failure in the silicone
joint, d) Test results highlighting range of linear material properties e) Volumetric finite element model of shoe connection f) Numerical analysis
response showing linear behaviour of silicone joint
 Manchester Town Hall, A case study in structural glass reliability and robustness
6.3. Quality Control Procedures.
A stringent quality control specification was defined to ensure the glass material strength was as expected. Small
scale fragmentation and mechanical strength tests were carried out to EN 12150 (2004). Tests were performed to
ensure surface compression met the minimum requirements for fully tempered glass as defined in BS EN 14179
(2005). Prior to testing, surface residual stresses were also measured using the scattered light polariscopic (SCALP)
procedure.

Full-scale fragmentation tests were carried out on the bent glass panels. Panels were fractured using an impactor,
and the fragmentation size recorded. Quality control tests were performed for each batch of panels manufactured,
with non-conforming panels rejected.

7. Conclusions
In structural glass design a simple structure is more deterministic; load paths can be clearly understood and the
requirement for complex modelling is reduced. Complex structures constructed of many components of varying
stiffnesses are difficult to analyse globally, and load paths may be incorrectly interpreted.

Simple glass structures however have an increased requirement for robustness, and post-failure capacity. If only one
load-path exists, there is a greater requirement to ensure a residual capacity in the event of failure. Careful analysis
of the load-redistribution after glass failure must be performed, and multiple failure modes considered.

In the absence of codified rules for the design of structural glass elements, an emphasis must be placed on ensuring
the design meets the minimum reliability requirements of BS EN 1990, and that the material assumptions used
within the reliability analysis are confirmed by quality control testing procedures.

Structural glass design is still a very young industry. The complexity of design is not always apparent when
reviewing a given scheme. It is important for designers to consider the load paths at every scale of the project, from
an initial global analysis level through to small scale connections. Although more and more guidance is available, it
is important for designers to not impart on the design of schemes beyond their understanding. In addition to a good
understanding of engineering principles, designers need an understanding quality assurance procedures, robustness
requirements, reliability analysis and validation of analysis procedures by project defined testing.

References
ASTM Standards (2012) E1300-12a, Standard practice for Determining Load Resistance of Glass in Buildings. West Conshohocken, PA: ASTM
International.
Bondi, S., McCllelland, N., Capturing Structural Silicone Non-Linear Behavior via the Finite Element Method, In: Proceedings of Glass
Performance Days, 2009, Finland
CEN, prEN 13474-3 (2008) Glass in building - Determination of the strength of glass panes - Part 3: General method of calculation and
determination of strength of glass by testing
CEN, BS EN 1990 (2002) Basis of Structural Design
Dias, V., Hechler, O., Odenbreit, C: Determination of Adhesive Properties for Non-linear Numerical Simulation of Structural Steel-Glass
Connections, In: Proceedings of Challenging Glass 3 – Conference on Architectural and Structural Applications of Glass, 2012 Delft, IOS
Press, pp. 195-207
Faber, M.H., Sørensen, J.D. CodeCal Software Version 03.01, (2003)
Institution of Structural Engineers, Structural use of Glass in Buildings 2014, The Institution of Structural Engineers, UK
Challenging Glass 5